Background: One-third of the US population experiences sleep loss, with the potential to impair physical and cognitive performance, reduce productivity, and imperil safety during work and daily activities. Computer-based fatigue-management systems with the ability to predict the effects of sleep schedules on alertness and identify safe and effective caffeine interventions that maximize its stimulating benefits could help mitigate cognitive impairment due to limited sleep. To provide these capabilities to broad communities, we previously released 2B-Alert Web, a publicly available tool for predicting the average alertness level of a group of individuals as a function of time of day, sleep history, and caffeine consumption.
Objective: In this study, we aim to enhance the capability of the 2B-Alert Web tool by providing the means for it to automatically recommend safe and effective caffeine interventions (time and dose) that lead to optimal alertness levels at user-specified times under any sleep-loss condition.
Methods: We incorporated a recently developed caffeine-optimization algorithm into the predictive models of the original 2B-Alert Web tool, allowing the system to search for and identify viable caffeine interventions that result in user-specified alertness levels at desired times of the day. To assess the potential benefits of this new capability, we simulated four sleep-deprivation conditions (sustained operations, restricted sleep with morning or evening shift, and night shift with daytime sleep) and compared the alertness levels resulting from the algorithm’s recommendations with those based on the US Army caffeine-countermeasure guidelines. In addition, we enhanced the usability of the tool by adopting a drag-and-drop graphical interface for the creation of sleep and caffeine schedules.
Results: For the 4 simulated conditions, the 2B-Alert Web–proposed interventions increased mean alertness by 36% to 94% and decreased peak alertness impairment by 31% to 71% while using equivalent or smaller doses of caffeine as the corresponding US Army guidelines.
Conclusions: The enhanced capability of this evidence-based, publicly available tool increases the efficiency by which diverse communities of users can identify safe and effective caffeine interventions to mitigate the effects of sleep loss in the design of research studies and work and rest schedules.
Previously, we developed and publicly released the 2B-Alert Web application , allowing users to compare and contrast predictions of alertness levels based on the psychomotor vigilance test (PVT) for a group of individuals as a function of time of day, sleep and wake schedule, and caffeine dose [ ]. Over the last 15 years, our group at the US Army incrementally developed and enhanced mathematical models that form the core of the 2B-Alert Web tool. At each developmental step, we created models with additional capabilities and independently validated the model predictions using an array of studies that investigated different sleep-deprivation conditions [ - ]. In total, we have assessed the model predictions for nearly 1200 participants from more than 50 laboratory and field studies. The study conditions ranged from sleep extension (10 h [hours] in bed) to chronic sleep restriction (3-7 h of sleep per night) to total sleep deprivation (TSD; up to 88 h of continuous wakefulness), with several studies performed to investigate the recuperative effects of caffeine administered in both single and multiple doses (50-600 mg), under a variety of sleep and wake schedules.
To date, the 2B-Alert Web tool has nearly 25,000 registered users from 144 countries, with more than 1800 users from 54 countries accessing the site ≥2 times in the last 12 months and a daily average of 26 log-ins. This is the only publicly available tool of its kind. Here, we describe an enhanced version of the tool, which has the added capability of automatically suggesting caffeine interventions (time and dose) to optimize alertness levels for desired (user-specified) times of day. For any particular combination of user-defined sleep and wake schedule, desired peak-alertness periods, maximum alertness-impairment threshold during these periods, and maximum total caffeine consumption in a 24-h period, the updated tool generates an optimal and safe caffeine-dosing schedule. Specifically, the algorithm generates schedules that use the least amount of caffeine to achieve a user-defined alertness level or achieve maximum alertness levels for a user-defined amount of caffeine . This capability would be particularly important during sleep-deprivation conditions because it maximizes the utility of caffeine as a fatigue countermeasure.
To attain this functionality, we incorporated a recently developed caffeine-optimization algorithm  with the predictive models [ , ] of the original 2B-Alert Web application [ ]. Previously, the tool predicted alertness levels for user-defined sleep and wake and caffeine schedules but required multiple trial-and-error simulations when the user wanted to determine a caffeine schedule that resulted in peak alertness levels during desired wake periods. In the new, enhanced version, this process is performed automatically, leading to a more effective means to identify safe caffeine interventions to guide the design of work and rest schedules and caffeine studies. Using this new capability of the tool, we demonstrated that, compared with the US Army guidelines for the use of caffeine as a countermeasure to sleep deprivation [ ], the 2B-Alert Web–proposed interventions increased mean alertness by 36% to 94% and decreased peak impairment by 31% to 71% while using equivalent or smaller doses of caffeine. In addition, we enhanced the usability of the tool by adopting a drag-and-drop graphical interface for the creation of sleep and caffeine schedules.
Original and Updated Capabilities
The enhanced 2B-Alert Web tool predicts PVT performance, a measure of alertness and sustained attention, for a typical individual (as determined from group data) as a function of time of day, sleep and wake schedule, and caffeine consumption (dose and time). It offers the capability to accomplish the following:
- Compare and contrast the effects of different sleep and wake and caffeine schedules on alertness.
- Automatically identify caffeine interventions that lead to the desired alertness levels at the desired times of day.
For the former capability, which is also offered in the original tool , it provides alertness-level (ie, PVT) predictions for a given sleep and wake and caffeine schedule ( ). Now, the updated tool offers the optional capability to automatically identify the optimal time and dose of caffeine consumption that lead to the desired, user-defined alertness outcomes. For (1) a specified sleep and wake schedule, (2) desired periods of peak alertness, (3) maximum acceptable alertness-impairment threshold during peak-alertness periods, and (4) maximum total caffeine consumption over a 24-h running period, the tool provides caffeine timing and dosing suggestions to achieve peak alertness at the desired times to the extent allowed by the limit on caffeine consumption ( ).
Overview of the 2B-Alert Web Tool
Upon user log-in, the tool provides brief stepwise instructions on how to select, edit, and add sleep and caffeine episodes to a schedule and make predictions (, right-hand panel). Similarly, when the user navigates to the caffeine optimization screen by clicking the Obtain Optimal Caffeine Schedule checkbox at the bottom of the right-hand panel in , the tool also provides brief stepwise instructions on how to create peak alertness periods, obtain an optimal caffeine schedule, and how to edit the alertness-impairment threshold and the maximum total caffeine consumption over a 24-h running period ( , right-hand panel). In addition, users can obtain brief descriptions of the functionalities of the various interaction-enabled buttons of the graphical interface by hovering over the buttons with a mouse. When the user clicks on the Help button in the upper right-hand corner of the display window, the system displays a user guide, which provides a comprehensive description of the functionalities of the system to facilitate assimilation of the 2B-Alert Web tool.
Comparing and Contrasting Different Sleep and Caffeine Schedules
The 2B-Alert Web tool can be used to compare and contrast the effects of different schedules on alertness. A schedule is defined by a series of sleep episodes and caffeine episodes spanning a period of up to 30 days (note that all day and time entries are for the same time zone). In turn, a sleep episode is defined by start and end days and times, whereas a caffeine episode is defined by the day, time, and dose of caffeine, with the dose entered manually or selected from a drop-down list of more than 30 popular caffeine-containing products. Upon user log-in, the tool displays three preloaded schedules: 8 h per night sleep, 5 h per night sleep, and 5 h per night sleep + 100 mg caffeine, with the middle schedule (yellow background) selected, as shown in the screenshot of the 2B-Alert Web tool interface ().
The user can modify an existing schedule or create a new schedule. For an existing schedule, users may view and edit the day and time of sleep episodes (gray icons) as well as caffeine-dose episodes (cup icons) by clicking on the corresponding icon. For example, when the user clicks on the sleep icon of the selected schedule in(5 h per night sleep), a pop-up window displays the start and end days of the sleep episode, along with the start (01:00) and end (06:00) times. Users may also add sleep and caffeine episodes by dragging and dropping the corresponding icon located above the schedules onto the timeline of the selected schedule. Users may select a schedule by clicking on its timeline or by choosing it from the drop-down list of preloaded timelines under Schedules on the right-hand side of the figure. The four buttons, from left to right, on the upper right-hand side of each schedule timeline allow users to export the schedule as a PDF file, change the schedule name, save a schedule with the same or a different name, or delete the schedule, respectively.
To create a new schedule, the user clicks the plus button to the right of the Schedule drop-down menu (, right-hand panel) and then adds sleep and caffeine episodes. Alternatively, a schedule can be imported from a Microsoft Excel file ( , Import Data) using a predefined format. Each generated or edited schedule can be saved with a user-defined name and exported as a PDF file ( , left-most of the 4 buttons on the upper right-hand side of this schedule’s timeline). The system supports up to 5 schedules and plots per session, where users can hide (or unhide) plots from view by clicking on the corresponding schedule name below the x-axis scroll bar. If the interface is loaded with 5 schedules, to create a new schedule, the user should delete an existing one.
The predicted alertness impairment for each schedule can be compared in the graph below the schedules.shows the 3 corresponding predictions for the 3 preloaded schedules using the mean response time (RT; in ms) PVT statistic, from day 0 through day 4 of the 4-day schedule, which starts on day 1 and lasts until the start of the sleep episode on day 5. The plots allow for the comparison of the effects of different sleep durations on alertness (eg, 5 h vs 8 h of sleep per night), as well as for the assessment of the beneficial effects of caffeine countermeasures (eg, 5 h per night of sleep with and without 100 mg of daily caffeine at 08:00). The displayed plots can be saved in an image file using the Export Graph button, and the numerical values for each of the 3 predicted statistics can be exported into an Excel file, along with the corresponding sleep and wake and caffeine schedules ( , Export Data). This allows users to import and reuse schedules in a future session because the system does not save schedules or their predictions on the web (the system erases all data when the user logs out).
In addition to the mean RT, the user can select to display the plots of alertness-impairment predictions for two other PVT statistics: mean speed (average of the reciprocal of RT; in 1/s) and number of lapses (number of RTs >500 ms). The statistics are for a 10-minute test and are selected from the Alertness Statistic drop-down menu located above the schedules (). To map PVT statistics into a more broadly understood metric of vigilance deficits, we used the findings from Dawson and Reid [ ] and Williamson et al [ ] to obtain an equivalence between PVT alertness-impairment values and blood alcohol concentrations (BACs). We estimated that a mean RT of 339 ms attained after 19 h of continued wakefulness corresponded to a 0.06% BAC ( , horizontal dashed gray line) and that a mean RT of 458 ms attained after 24 h of continuous wakefulness corresponded to a 0.08% BAC. For context, driving at BACs of 0.06% and 0.08% increase the risk of causing a traffic accident by 2- and 3-fold, respectively, compared with control drivers [ , ].
Automatic Identification of Caffeine Interventions
To automatically identify caffeine interventions that lead to desired alertness levels for a selected sleep and wake schedule, the user should click the Obtain Optimal Caffeine Schedule checkbox at the bottom of the right-hand panel in, which takes users to another graphical interface ( ). In this interface, the user can add a desired period of peak alertness (start and end days and times), which is a required input, to the schedule. This is achieved by dragging and dropping the Peak Alertness icon (green icon) located above the schedules onto the timeline ( ). Next, clicking the Optimize button in the lower right-hand panel will generate the optimal caffeine schedule and display the corresponding alertness-impairment prediction. For the selected schedule shown in (5 h per night sleep), these steps resulted in the identification of the optimal caffeine intervention named 5 h per night Sleep + Optimal caffeine solution in and the corresponding alertness-impairment prediction plot. In this case, peak-alertness periods from 08:00 to 16:00 for days 1 to 4 required a single caffeine dose at 07:00 of 100 mg on day 1 and 200 mg on day 2 and two doses at 07:00 and 10:00 of 200 mg and 100 mg, respectively, on days 3 and 4 (see Figure S1 in for details; larger cup icons indicate larger caffeine doses). Optionally, the Advanced Options checkbox at the bottom of the lower right-hand panel allows users to set thresholds for maximum total caffeine in a 24-h period (100-1500 mg; default 400 mg) and the maximum alertness-impairment level (RT ranging from 150-500 ms; default 300 ms) by dragging the sliders to the desired values ( ). As described previously, users can rename, save, delete, and export the updated schedule, as well as save the displayed plot as an image file. Unchecking the Obtain Optimal Caffeine Schedule in the upper right-hand panel takes users back to the prediction graphical interface ( ).
Initial Conditions and Assumptions
We formulated the predictive model in the 2B-Alert Web tool so that alertness is an inversely related function of sleep debt, which accumulates over days with <8 h of sleep per day [, , ] but decreases for sleep durations of >8 h per day. We initialized the tool so that on day 0 there is no sleep debt after 8 h of sleep (23:00-07:00). However, if sleep debt is nonzero, users need to enter up to 7 days of sleep history at the beginning of the schedule (sleep episodes >7 days old have negligible influence on near-future alertness [ , , ]).
We assumed that the restorative effect of caffeine depends on the alertness level; that is, for a given caffeine dose, the larger the alertness impairment, the greater the beneficial effect of caffeine [, ], where the magnitude of the benefit depends on the impairment level, the caffeine dose, and the residual concentration of caffeine from previous doses.
For the automatic identification of caffeine interventions that optimize alertness levels, we formulated the optimization problem so that, for the desired periods of peak alertness, the algorithm seeks caffeine-intervention solutions that equally weigh the cumulative deficit above the maximum alertness-impairment threshold and the peak alertness-impairment level . To obtain practical and safe solutions, we added the following constraints: (1) caffeine doses to be restricted to 100 mg, 200 mg, or 300 mg; (2) dosing to occur on the hour; (3) the minimum time period between doses to be 2 h; and (4) the cumulative caffeine concentration in the blood to be below the maximum level achieved by a single 400-mg dose [ ]. The optimization algorithm finds solutions in a matter of seconds by generating and assessing only caffeine schedules that are likely to reduce alertness impairment [ ]. The tool attempts to find solutions that use the least amount of caffeine, while meeting the imposed constraints. However, in certain cases, even when using the maximum amount of caffeine, it may not be possible to obtain solutions that reduce the alertness impairment below the maximum impairment threshold. To obtain solutions with lower alertness impairment, users may increase the maximum total caffeine consumption in a 24-h period ( , Advanced Options). Alternatively, the near-optimal 2B-Alert Web solution provides a good starting point for manual exploration of additional caffeine interventions that are potentially more effective but likely violate specified constraints.
Simulations to Assess the Optimization Algorithm
To assess the efficacy and potential benefits of the automated caffeine interventions proposed by the 2B-Alert Web algorithm, we performed 4 simulations encompassing the three circumstances under which caffeine countermeasures are considered by the US Army guidelines : sustained operations, restricted sleep, and night shift. summarizes the US Army guidelines for the use of caffeine as a countermeasure to sleep deprivation and the simulated conditions, including the sleep-deprivation schedules and the assumed periods of peak alertness. For example, for the sustained operations scenario in condition 1, we simulated a 30-h TSD challenge from 07:00 on day1 to 13:00 on day2 and arbitrarily assumed a 13-h peak-alertness period from 00:00 to 13:00 on day2 and a maximum alertness-impairment threshold of 270 ms ( ), which corresponds to the maximum alertness impairment under well-rested conditions (ie, habitual sleep of 8 h per day). Then, we used the tool to obtain an optimal caffeine schedule and compared the resulting alertness impairment with that obtained by using the US Army guidelines. Figures S2-S7 in provide details for the simulations of conditions 2-4.
|Condition||Army caffeine guideline||Sleep schedule||Peak-alertness period||Countermeasure recommendation|
|1. Sustained operations: 30 ha of total sleep deprivation||200 mg every 4 h as needed, starting at 00:00||N/Ab||00:00-13:00 on day 2||800 mg in 4 doses (A)||800 mg in 4 doses (B)|
|2. Chronic sleep restriction: morning shift||200 mg upon awakening, 200 mg 4 h later||01:00-06:00 for 5 days||08:00-16:00 for 5 days||2000 mg in 10 doses (Figure S3Ac)||1900 mg in 13 doses (Figure S3Bc)|
|3. Chronic sleep restriction: evening shift||200 mg upon awakening, 200 mg 4 h later||01:00-06:00 for 5 days||15:00-23:00 for 5 days||2000 mg in 10 doses (Figure S5Ac)||1700 mg in 11 doses (Figure S5Bc)|
|4. Night shift with daytime sleep||200 mg at the beginning of the shift||20:00 to 22:00 and 10:00 to 15:00 for 5 days||00:00 to 08:00 for 5 days||1000 mg in 5 doses (Figure S7Ac)||1000 mg in 5 doses (Figure S7Bc)|
bN/A: not applicable.
cSee Figures S3, S5, and S7 in.
Alertness Improvement Metrics
To quantify the potential benefit of the caffeine-optimization algorithm, we used two metrics: mean alertness impairment and peak alertness impairment, each computed over the desired periods of peak alertness. The mean alertness impairment accounts for the overall impairment level across the peak alertness periods, whereas the peak alertness period captures acute levels of impairment over short periods of time during the peak alertness periods. For the mean RT PVT metric used in our simulations, we defined the mean alertness impairment (in ms) as the ratio of the area of the predicted alertness impairment above the selected maximum alertness-impairment threshold to the duration of the peak alertness periods. Similarly, we defined the peak alertness impairment (in ms) as the maximum mean RT over all periods of desired peak alertness minus the selected maximum alertness-impairment threshold. In addition, we computed the total time (in h) that the predicted alertness impairment was above the 0.06% BAC threshold during the peak alertness periods as a measure of time for which an individual is at an increased risk of making a mistake because of deficits in alertness.
The overall software architecture of the updated 2B-Alert Web tool was unchanged from the original release except for the overhaul of the graphical user interface and the use of updated versions of the underlying software technologies. We hosted the tool on an Apache Tomcat web server and provided access through a secure service over HTTP Secure. We used a three-tier architecture consisting of (1) a PostgreSQL database server, which stores user account information; (2) a controller, which provides access to the alertness-prediction model and caffeine-optimization algorithm and implements the functionalities required to create and manage multiple predictions and optimizations; and (3) an interaction-enabled user interface, which provides the ability to create schedules by dragging and dropping sleep, caffeine, and peak-alertness episodes onto the schedule timeline; show and hide plots; and dynamically zoom into and out of plots. The system runs without any plugins and is accessible through multiple web browsers, including Internet Explorer version 11, Chrome version 74, and Firefox version 67, or earlier versions thereof.
Access and Privacy
The 2B-Alert Web tool is freely available to registered users through a secure web browser. Registration consists of users providing their name, email address, and affiliation, after which they receive a confirmation email with a username and password, which can be changed. The system erases all simulated schedules and results when the user logs out, providing user privacy and maintaining data confidentiality.
Potential Benefits of the 2B-Alert Web Caffeine-Optimization Algorithm
To demonstrate the potential benefit of the 2B-Alert Web caffeine-optimization algorithm, we compared the effects of caffeine countermeasures as recommended by the US Army guidelines  with those of the tool for the 4 conditions illustrated in . For the sustained-operations scenario in condition 1 ( ), the guideline recommended an initial 200-mg caffeine dose at midnight, followed by 200 mg every 4 h thereafter (at 04:00, 08:00, and 12:00) for a total of 800 mg over a 12-h span ( A). shows the resulting alertness-impairment prediction plot in terms of the mean RT (red line). also shows the optimal caffeine schedule recommended by the tool and the associated alertness-impairment prediction plot (green line) for the same total amount of 800 mg of caffeine. B, obtained by exporting the Optimal caffeine solution for 30 h of TSD schedule as a PDF, shows the numerical values of the optimal caffeine times and doses recommended by the tool. By suggesting caffeine interventions at different time intervals in varying doses of 100 mg, 200 mg, or 300 mg over a more compressed 7-h span during which time the circadian-mediated alertness deficits were greatest, the 2B-Alert Web solution consistently kept impairment below the 0.06% BAC equivalence level during the entire peak-alertness period. In contrast, the US Army guideline solution was less effective, resulting in 3 time periods when alertness deficits exceeded the 0.06% BAC level for a total of 3.4 h above this threshold and impairment levels as high as 378 ms (270+108 ms) around 08:00 on day 2 ( and ). To provide a context of the significance of this difference, an individual following the US Army guidelines would have a 2-fold increased risk of causing a traffic accident during the 3.4-h period compared with substantially lower increase in risk with the tool’s recommendations. Nevertheless, for a maximum of 800 mg of caffeine, neither recommendation was capable of maintaining alertness levels below the maximum user-defined threshold of 270 ms for most of the 13-h period of peak alertness.
|Condition||Mean alertness impairment (ms)||Improvement (%)||Peak alertness impairment (ms)||Improvement (%)||Time above 0.06% BAC equivalent (h)|
|Army||2B-Alert Web||Army||2B-Alert Web||Army||2B-Alert Web|
Similarly, the 2B-Alert Web caffeine recommendations for the sleep restriction conditions and the night shift with daytime sleep condition inyielded superior results when compared with the US Army guidelines solutions. The tool’s recommendations yielded larger improvements in mean alertness impairment and peak alertness impairment, while using the same or as much as 300 mg less caffeine than the US Army’s recommendations ( and ). In condition 2, although the percentage improvement in mean alertness impairment was 50%, the absolute improvement averaged over the peak-alertness periods in the 5 simulated days was minimal (4 ms vs 8 ms). Given the limited amount of sleep deficit, the relatively narrow period of desired peak alertness (8 h per day), and the relatively large amount of consumed caffeine per day (400 mg), there were not enough degrees of freedom in this problem to demonstrate a considerable improvement of the optimization algorithm over the guidelines solution because it already provided near-optimal benefits. For related reasons, at times, the benefits achieved over peak-alertness periods came at the detriment of other wakefulness periods (eg, condition 3). Figures S2-S7 in provide detailed results for the simulations of conditions 2-4. Overall, our simulation results show that customized caffeine recommendations intended to enhance alertness during a specific time of the day consistently resulted in higher levels of alertness compared with suitable but generic recommendations.
We have enhanced the publicly available 2B-Alert Web tool to automatically identify caffeine interventions that safely optimize alertness within selected time periods for any user-specified sleep and wake schedule. Users can use the tool to compare and contrast the effects of different sleep and wake and caffeine schedules on the alertness of a group of individuals and to automatically identify safe caffeine interventions to achieve and sustain desired alertness levels for desired times of day. To achieve this added capability, we integrated a well-validated mathematical model, which predicts alertness impairment levels based on the PVT as a function of time of day, sleep and wake schedule, and caffeine consumption, with a recently developed caffeine-optimization algorithm, which identifies the minimum amount of caffeine required to restore alertness levels, while imposing multiple operational and safety constraints (some of which can be modified by the user). The 2B-Alert Web tool synthesizes decades of sleep-physiology and mathematical modeling research by the US Army into an open-access practical tool that should find value in research and operational communities.
To demonstrate the potential benefits of automatically identifying caffeine recommendations that safely optimize alertness for desired time periods, we simulated four conditions of military relevance reflective of sustained operations, restricted sleep (with morning and evening shifts), and night shift with daytime sleep (). Next, we compared the resulting alertness impairment levels by considering the US Army guidelines for the use of caffeine as a countermeasure to sleep deprivation [ ] against those automatically obtained by using the newly implemented caffeine optimization capability [ ] provided by the tool. The results show that the tool consistently provides solutions with greater improvements in mean alertness impairment (by 36% to 94%; mean 59%, SD 25%) and peak alertness impairment (by 31% to 71%; mean 45%, SD 18%), while using the same or as much as 300 mg less caffeine than the US Army’s recommendations ( ). Although the extent of such benefits varies as a function of sleep and wake schedule, time and duration of desired peak alertness periods, acceptable alertness-impairment threshold, and maximum total amount of caffeine consumption, the results also demonstrate that the customized caffeine recommendations provided by this new capability are superior to generic, one-size-fits-all guidelines.
The tool has limitations. First, the predictive models that form the core of the 2B-Alert Web application have only been validated for healthy, young adults of military age. The extent to which the tool is applicable to an older, heterogeneous cohort is not known. Second, the tool’s predictions of alertness impairment are in fact surrogates for PVT statistics. Hence, whether such predictions also serve as surrogates for other neurocognitive measures of performance impairment remains to be investigated. Third, all sleep and caffeine times are assumed to be for the same time zone. The ability to predict alertness impairment involving transmeridian travel would require circadian-phase data across time zones, which we do not possess. Fourth, the 2B-Alert Web tool does not account for the potential disruptive effect of caffeine when consumed soon before sleep episodes. In a future implementation, we could address this limitation by forbidding caffeine recommendations during certain times of the day. Fifth, the predictive models do not account for tolerance to caffeine. However, a sensitivity analysis of the caffeine-model component suggests that changes in model parameters to account for this effect have a very small impact on the predicted alertness . Importantly, experimental observations indicate that the main determinant of the effect of caffeine on alertness is the impairment level, regardless of the caffeine tolerance level [ , ], which is consistent with the model predictions.
Sixth, the tool has not been prospectively tested in military or civilian settings such as the Marine Corps Officer Basic School or shiftwork at hospitals, and we estimated the benefits of the optimal caffeine schedules for the 4 simulation scenarios inand using the tool’s prediction models instead of using actual PVT measurements from an experimental study. However, the mathematical models at the core of the tool have been thoroughly validated in dozens of different sleep and caffeine-consumption conditions. For example, in one of the validation analyses, Ramakrishnan et al [ ] showed that approximately 87% of the model predictions are within 2 SEs of the measured mean data, meaning that, 87% of the time, the model group-average predictions are statistically indistinguishable from the experimental data. Thus, we have reason to believe that the model will provide good predictions in conditions for which experimental data are not available, producing adequate guidance for sleep and caffeine schedules when it is not possible or desirable to conduct a study. Finally, we limited predictions to a maximum of 30 consecutive days to be able to provide solutions in a few (less than approximately 3) minutes. Nevertheless, the user can run longer scheduling scenarios by carefully breaking the problem into smaller chunks.
The new and unique capability to automatically identify caffeine interventions that safely optimize alertness transforms the 2B-Alert Web tool from an alertness and performance prediction tool into an alertness and performance enhancement tool. This functionality not only results in peak alertness levels during desired wake periods but, by recommending the right caffeine doses at correct times, also maximizes the utility of caffeine. The 2B-Alert Web tool provides caffeine schedules that use the least amount of caffeine to achieve a desired alertness level, thus reducing potential side effects of excessive caffeine use, or that yield maximum alertness levels for a given amount of caffeine, thus optimizing its recuperative effects.
The authors express their gratitude to Dr Tatsuya Oyama and Ms Maria Kuhrmann for editing the manuscript, Dr Bora Sul for providing constructive feedback, and the hundreds of study volunteers whose data we used to construct and validate our mathematical models over the years. This work is supported by the US Army Medical Research and Development Command under contract number W81XWH20C0031. This was not an industry-supported study. This work was sponsored by the Military Operational Medicine Program Area Directorate of the US Army Medical Research and Development Command, Fort Detrick, MD. The opinions and assertions contained herein are the private views of the authors and are not to be construed as official or as reflecting the views of the US Army, the US Department of Defense, or the Henry M. Jackson Foundation for the Advancement of Military Medicine, Inc. This paper has been approved for public release with unlimited distribution.
JR, SR, and FGVL conceived the research. KK, LH, AF, and FGVL developed the software and performed the simulations. TJD and TJB provided user-perspective input to software workflow and graphical user interface. JR and FGVL wrote the manuscript, which was reviewed and edited by all authors.
Conflicts of Interest
JR, TJB and FGVL receive royalties for the licensing of the 2B-Alert technology from Integrated Safety Support.
Supporting information.DOC File , 863 KB
- Biotechnology High Performance Computing Software Applications Institute. URL: https://2b-alert-web.bhsai.org [accessed 2021-01-13]
- Reifman J, Kumar K, Wesensten NJ, Tountas NA, Balkin TJ, Ramakrishnan S. 2B-Alert Web: an open-access tool for predicting the effects of sleep/wake schedules and caffeine consumption on neurobehavioral performance. Sleep 2016 Dec 01;39(12):2157-2159 [FREE Full text] [CrossRef] [Medline]
- Rajaraman S, Gribok AV, Wesensten NJ, Balkin TJ, Reifman J. Individualized performance prediction of sleep-deprived individuals with the two-process model. J Appl Physiol (1985) 2008 Feb;104(2):459-468 [FREE Full text] [CrossRef] [Medline]
- Rajaraman S, Gribok AV, Wesensten NJ, Balkin TJ, Reifman J. An improved methodology for individualized performance prediction of sleep-deprived individuals with the two-process model. Sleep 2009 Oct;32(10):1377-1392 [FREE Full text] [CrossRef] [Medline]
- Rajdev P, Thorsley D, Rajaraman S, Rupp TL, Wesensten NJ, Balkin TJ, et al. A unified mathematical model to quantify performance impairment for both chronic sleep restriction and total sleep deprivation. J Theor Biol 2013 Aug 21;331:66-77 [FREE Full text] [CrossRef] [Medline]
- Ramakrishnan S, Laxminarayan S, Wesensten NJ, Kamimori GH, Balkin TJ, Reifman J. Dose-dependent model of caffeine effects on human vigilance during total sleep deprivation. J Theor Biol 2014 Oct 07;358:11-24. [CrossRef] [Medline]
- Ramakrishnan S, Wesensten NJ, Balkin TJ, Reifman J. A unified model of performance: validation of its predictions across different sleep/wake schedules. Sleep 2016 Jan 01;39(1):249-262 [FREE Full text] [CrossRef] [Medline]
- Ramakrishnan S, Wesensten NJ, Kamimori GH, Moon JE, Balkin TJ, Reifman J. A unified model of performance for predicting the effects of sleep and caffeine. Sleep 2016 Oct 01;39(10):1827-1841 [FREE Full text] [CrossRef] [Medline]
- Vital-Lopez FG, Ramakrishnan S, Doty TJ, Balkin TJ, Reifman J. Caffeine dosing strategies to optimize alertness during sleep loss. J Sleep Res 2018 Oct;27(5):e12711 [FREE Full text] [CrossRef] [Medline]
- Army Techniques Publication ATP 6-22.5 A Leader’s Guide To Soldier Health And Fitness. United States: Army Techniques Publication; 2016.
- Dawson D, Reid K. Fatigue, alcohol and performance impairment. Nature 1997 Jul 17;388(6639):235. [CrossRef] [Medline]
- Williamson AM, Feyer AM, Mattick RP, Friswell R, Finlay-Brown S. Developing measures of fatigue using an alcohol comparison to validate the effects of fatigue on performance. Accid Anal Prev 2001 May;33(3):313-326. [CrossRef] [Medline]
- Borkenstein R. The Role of the Drinking Driver in Traffic Accidents (the Grand Rapids Study). Bloomington: Indiana University; 1974.
- Martin TL, Solbeck PA, Mayers DJ, Langille RM, Buczek Y, Pelletier MR. A review of alcohol-impaired driving: the role of blood alcohol concentration and complexity of the driving task. J Forensic Sci 2013 Sep;58(5):1238-1250. [CrossRef] [Medline]
- Landolt H, Rétey JV, Adam M. Reduced neurobehavioral impairment from sleep deprivation in older adults: contribution of adenosinergic mechanisms. Front Neurol 2012;3:62 [FREE Full text] [CrossRef] [Medline]
- Rétey JV, Adam M, Gottselig JM, Khatami R, Dürr R, Achermann P, et al. Adenosinergic mechanisms contribute to individual differences in sleep deprivation-induced changes in neurobehavioral function and brain rhythmic activity. J Neurosci 2006 Oct 11;26(41):10472-10479 [FREE Full text] [CrossRef] [Medline]
- Caffeine for the Sustainment of Mental Task Performance Formulations for Military Operations. Washington, DC: The National Academies Press; 2001.
|BAC: blood alcohol concentration|
|PVT: psychomotor vigilance test|
|RT: response time|
|TSD: total sleep deprivation|
Edited by R Kukafka; submitted 14.04.21; peer-reviewed by R Marshall, YH Lin, YS Huang; comments to author 24.09.21; revised version received 12.10.21; accepted 15.11.21; published 27.01.22Copyright
©Jaques Reifman, Kamal Kumar, Luke Hartman, Andrew Frock, Tracy J Doty, Thomas J Balkin, Sridhar Ramakrishnan, Francisco G Vital-Lopez. Originally published in the Journal of Medical Internet Research (https://www.jmir.org), 27.01.2022.
This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on https://www.jmir.org/, as well as this copyright and license information must be included.